Math problem with results, page-261

AICp ..... I think I am the only one that sees your point ... but it is not quite as black and white as you are presenting it.

Yes ... you are correct ... double blinding becomes less a potential source of bias when outcomes are “hard”.

But double blinding in a diagnostic accuracy study acts as an insurance policy against “contamination” between the assessors of the index test (the new test) and the assessors of the reference test knowing what each other have arrived at.

In the STARD reporting framework for diagnostic testing the term “double blinding” isn’t actually used. Rather it is just a matter of reporting whether index testers knew what the clinical team were doing and visa versa. This reflects I think this subtle difference between double blinding for controlling bias (placebo effects etc) and double blinding to prevent contamination.

But where doctorbrad is too black and white is that while in clinical trials blinding is always crystal clear (because it is such a important source of bias) in diagnostic accuracy trials often blinding is not considered a very important issue and isn’t reported.

But this is a side issue for the real difference between the WA study and US trial. Diagnostic accuracy testing is a “study” when it uses retrospective data. Diagnostic accuracy testing trips over to become a “trial” when the data is collected prospectively. This is the critical difference between the two; not the blinding.